Many countries in Africa now hope to provide highly active antiretroviral therapy (ART) to people with AIDS. A major concern is that most health facilities in Africa have shortages of clinical staff and are poorly equipped. Travel to health facilities, particularly hospitals, is often difficult because of poor transport infrastructure and the long distances involved. There is an urgent need to identify appropriate models of delivery for ART. We plan to conduct a phase III randomized clinical trial to compare ART delivery through two different models: a) ART delivered through health facilities by clinically qualified staff and b) home-based care in which non-medically qualified lay workers play a major role in the ART delivery and clients are followed up at health facilities much less frequently. The primary objective will be to measure the effects of these strategies on plasma viral load approximately 36 months after randomization. We will also examine the effects on treatment failure, adherence, and describe the effects of the introduction of ART on the health service. The utility of different methods of measuring adherence, the factors associated with poor adherence and changes in disclosure behavior that may occur over time among clients receiving ART, will be explored. Randomization will be done by clusters, defined using sub-counties in the district, and stratified by distance from fixed health facilities, and urban/rural. Just over 800 participants, living in 40 clusters, will be recruited over a period of 3-6 months and follow-up over a period of 3 years. Assuming a coefficient of variation of 0.2, and a loss-to-follow-up of 15%, and a rate of detectable plasma viral load of 30% over the 3 years, the trial will have 80% power to show that the upper limit of the 95% two-sided confidence interval of the difference in the rate of detectable plasma viral load between the two arms does not exceed 10% (i.e. that one of the arms is at worst 10% inferior to another).